Xuewu Lin

Xuewu Lin

Auto-generated profile, based on publicly available information

Xuewu Lin

Horizon

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
13 October 2023 in IEEE Sensors Journal
Verified
Journal Article
Closed access
YGYu-Ang Guo
Yu-Ang Guo
JWJianqiang Wang
Jianqiang Wang
GYGuizhen Yu
Guizhen Yu
CZChuang Zhang
Chuang Zhang
XLXuewu Lin
Xuewu Lin
GSGaolei Shi
Gaolei Shi
The recognition of road weather conditions can provide environmental information for the autonomous driving systems, enabling them to develop safe driving strategies. Previous studies have utilized cameras or radars to identify road weather conditions, but they were unable to detect specific dry, wet, or snowy road areas. Some methods based on image segmentation can detect these areas, but their performance is limited by the similar features exhibited by dry and wet roads. In this paper, we propose a novel method for segmenting dry, wet, and snowy road areas based on the fusion of images and point cloud reflection intensities. By fusing the point cloud reflection intensity, we enhance the distinction between different road types, thereby improving the accuracy of the segmentation. Furthermore, we address the challenge posed by sparser point cloud data compared to image pixels by introducing a point cloud reflection intensity upsampling method. Unlike other interpolation methods, our proposed method relies on the physical principle that an object’s reflectivity is correlated with its color and texture features, which makes the point cloud reflection intensity obtained by upsampling more reliable. To evaluate the performance of our method, we conducted experiments on a self-collected dataset that contains over 20,000 images. More than 4,000 images were annotated at the pixel level, and each of them was accompanied by corresponding point cloud data. The experimental results demonstrate that our proposed algorithm achieves an impressive mIoU (mean Intersection over Union) of 87.94%. Compared to image segmentation algorithms that do not fuse point clouds, our algorithm achieves an approximately 3% improvement in mIoU, substantiating its effectiveness.

4 June 2026 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
LLLongzhong Lin
Longzhong Lin
XLXuewu Lin
Xuewu Lin
KXKechun Xu
Kechun Xu
Haojian LUHaojian LU
Haojian LU
LHLichao Huang
Lichao Huang
Yue WangYue Wang
Yue Wang
Simulation plays a crucial role in assessing autonomous driving systems, where the generation of realistic multi-agent behaviors is a key aspect. In multi-agent simulation, the primary challenges include behavioral multimodality and closed-loop distributional shifts. In this study, we formulate a unified mixture model (UniMM) framework revisit mixture models for generating multimodal agent behaviors, which can cover the mainstream methods including regression-based continuous mixture models and discrete NTP GPT-like discrete models. Furthermore, we introduce a closed-loop sample generation approach tailored for mixture models to mitigate distributional shifts. Within the UniMM framework, we recognize critical configurations from both the model and data perspectives. We conduct a systematic examination of various model configurations, and comprehensively characterize their effects including positive component matching, continuous regression, prediction horizon, and the number of components. Moreover, our investigation into the data configuration highlights the pivotal role of closed-loop samples in achieving realistic simulations. To extend the benefits of closed-loop samples across a broader range of mixture models, we further introduce a temporal disentanglement-and-alignment mechanism to address the shortcut learning and off-policy learning issues. Leveraging insights from our exploration, the distinct variants proposed within the UniMM framework, including discrete, anchor-free, and anchor-based models, all achieve state-of-the-art performance on the WOSAC benchmark.

1 November 2026 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
WSWenchao Sun
Wenchao Sun
XLXuewu Lin
Xuewu Lin
KCKeyu Chen
Keyu Chen
ZPZixiang Pei
Zixiang Pei
YSYining Shi
Yining Shi
SZSifa Zheng
Sifa Zheng